Re: layer 4 switch for efficient/scaleable cache use?

From: Eric Stern <estern@dont-contact.us>
Date: Thu, 4 Mar 1999 13:04:48 -0500 (EST)

On Wed, 3 Mar 1999, Jon Zeeff wrote:

> I've been unable to find a layer 4 http switch that really understands
> what is needed to be an efficient frontend for a cache farm (multiple
> cache boxes at the same place). Of course you want the same URLs to
> go to the same cache box (minimizing disk usage, icp traffic, etc).
> But you also want even load distribution, even in the case where there
> are so many hits to one site that a single cache box can't handle it.
> Ie, if www.acme.com is getting lots of hits, an efficient thing to do
> would be to send all of http://www.acme.com/[a-n].* to cache1 and
> http://www.acme.com/[m-z].* to cache2*. You cannot simple create some
> hash value based on the ip address or domain name www.acme.com. And
> of course, if cache1 goes down, its load needs to be distributed among the
> remaining boxes.
>
> Anyone know of layer 4 switches that do this? We need flexible and
> dynamic load distribution that always (or at least under steady state
> conditions) directs the same request to the same single cache.

Our TrafficCop product may be what you are looking for. (See our website
listed below). Its not exactly a layer 4 switch, but it IS designed to do
exactly what you describe. We use it as a redirector for our web caching
product. It uses the CARP protocol to distribute requests across your
caches.

Its not as high performance as say an Alteon ACEDirector or a Foundry
ServerIron, but then it doesn't cost near as much either. Deployed
properly (and they should be deployed a little differently than a typical
layer 4 switch) they are very effective.

/-----------------------------------------------------------------------/
/ Eric Stern - PacketStorm Technologies - (519) 826-9395 /
/ http://www.packetstorm.on.ca /
/ WebSpeed - a transparent web caching server - available now! /
/-----------------------------------------------------------------------/
Received on Thu Mar 04 1999 - 11:06:43 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:45:08 MST